An intelligent question and answer system based on converter station operation and maintenance

Abstract

The application provides an intelligent question and answer system based on converter station operation and maintenance, the system comprises: an intelligent interaction module: for receiving user input questions, and presenting the optimal answer generated by the answer generation module to the user according to the input question; an operation and maintenance knowledge module: for acquiring knowledge data related to converter station operation and maintenance, and constructing a converter station operation and maintenance knowledge graph; an intention understanding module: for performing word segmentation on the user input question, and analyzing the intention of the input question according to the word segmentation; an answer generation module: for receiving the question processed by the intention understanding module, and generating an optimal answer by reasoning the question in combination with the converter station operation and maintenance knowledge graph. The application can generate an optimal answer according to the user input question, realize intelligent answering and processing of converter station operation and maintenance related questions, solve the information fragmentation and dependence on artificial experience problems in the existing converter station operation and maintenance, and improve the efficiency and accuracy of operation and maintenance answering.

Description

Technical Field

[0001] This invention relates to the field of intelligent question-and-answer technology, and in particular to an intelligent question-and-answer system based on converter station operation and maintenance. Background Technology

[0002] As the core hub of a high-voltage direct current (HVDC) transmission system, converter stations undertake the critical tasks of power conversion and transmission, and their operational status directly affects the safety and stability of the power system. Converter stations involve a wide variety of equipment with complex structures, and their operation and maintenance work encompasses multiple aspects, including equipment monitoring, fault diagnosis, operational execution, and compliance with procedures, demanding extremely high levels of professional knowledge and experience from maintenance personnel.

[0003] Currently, during the operation and maintenance of converter stations, staff mainly rely on consulting paper technical manuals, electronic operation and maintenance procedures, senior experts, or historical operation and maintenance records to obtain solutions when encountering problems. This approach has drawbacks: First, operation and maintenance technical information is scattered across different systems and documents, resulting in severe information fragmentation and low retrieval efficiency; second, it relies too heavily on human experience, which can easily lead to judgment errors and makes it difficult to quickly obtain accurate solutions; third, existing question-and-answer systems are mostly based on general power knowledge, lacking specialization and specificity for converter station operation and maintenance scenarios, resulting in inaccurate semantic understanding and poor practicality of the question-and-answer results. Summary of the Invention

[0004] Therefore, the purpose of this invention is to provide an intelligent question-and-answer system based on converter station operation and maintenance, so as to solve or partially solve the above-mentioned problems existing in the prior art.

[0005] To achieve the above objectives, the present invention provides an intelligent question-and-answer system based on converter station operation and maintenance, the system comprising: Intelligent interaction module: Used to receive questions input by the user and present the best answer generated by the answer generation module based on the input question to the user; Operation and Maintenance Knowledge Module: Used to acquire knowledge data related to converter station operation and maintenance, and to build a converter station operation and maintenance knowledge graph; Intent understanding module: This module segments the user's input question into words and analyzes the intent behind the input question based on the word segmentation. Answer generation module: It receives questions processed by the intent understanding module, combines the questions with the converter station operation and maintenance knowledge graph to reason and generate the optimal answer.

[0006] Furthermore, the intelligent interaction module supports multiple interaction methods, including text input, voice input, and touch operation.

[0007] Furthermore, the intelligent interaction module is also used to output the optimal answer generated by the answer generation module in the form of text, charts and step-by-step instructions, and to record user question-and-answer interaction logs, which are used for the optimization of the intent understanding module.

[0008] Furthermore, the construction of the converter station operation and maintenance knowledge graph specifically includes the following steps: S11. Collect various knowledge data related to converter station operation and maintenance, including converter station equipment ledger data, converter station operation and maintenance technical manual, and fault case data, and perform preprocessing operations on the collected knowledge data. S12. Divide the preprocessed knowledge data into structured data, semi-structured data, and unstructured data. Extract the first operational features from the structured and semi-structured data. Perform word segmentation and feature word extraction on the unstructured data. S13. Combine and analyze the first operation and maintenance features extracted in step S12 with the unstructured data after word segmentation and feature word extraction to construct a converter station operation and maintenance dictionary; S14. Based on the converter station operation and maintenance dictionary, determine the converter station equipment categories and their characteristics in various knowledge data, label the converter station equipment categories and their characteristics, and integrate the labeled converter station equipment categories and their characteristics into a training dataset. S15. Construct an operation and maintenance knowledge model. Input the training dataset into the operation and maintenance knowledge model for training. Use the trained operation and maintenance knowledge model to extract the second operation and maintenance feature from the unstructured data. Merge the second operation and maintenance feature with the first operation and maintenance feature to construct the converter station operation and maintenance knowledge graph.

[0009] Furthermore, the operation and maintenance knowledge module is also used to fuse and process the acquired knowledge data, specifically including the following steps: S21. Based on the acquired knowledge data, extract the features of each entity in the knowledge data, and use the entity features to analyze the feature similarity between entities; S22. Based on the feature similarity between entities, match different entities to generate a potential matching set for entity matching; S23. Preset entity matching rules and filter the potential matching set according to the entity matching rules; S24. Based on the screening results of the potential matching set, establish matching relationships between different entities to eliminate knowledge redundancy in the converter station operation and maintenance knowledge graph.

[0010] Furthermore, the intent understanding module specifically includes the following steps: S31. Use lexical analysis to segment the user's input question into words or phrases; S32. Initialize the word segmentation obtained in step S31 to obtain multi-dimensional word segmentation vectors, and construct the initialization matrix of the problem based on the multi-dimensional word segmentation vectors. S33. Model the initialization matrix using convolution, extract local features in the problem through convolution operations, and capture the correlation between word segments; S34. Introduce a convolutional neural network model to process the initialization matrix after convolutional modeling, extract the user's intent features, and analyze the user's final intent based on the intent features.

[0011] Furthermore, the answer generation module specifically includes the following steps: S41. Based on the user's final intent, extract the entity from the question input by the user to obtain the question entity; S42. Match the problem entities with the converter station operation and maintenance knowledge graph to obtain the optimal candidate entity for each problem entity, and construct a problem entity set based on the optimal candidate entities; S43. Semantically vectorize the problem entity set and the converter station operation and maintenance knowledge graph, and perform semantic similarity matching between the user's final intent and the semantically vectorized problem entity set and knowledge graph to obtain a set of relationships; S44. Based on the user's final intent, the set of question entities, and the set of relationships, construct a mapping relationship with the knowledge graph, and deduce the optimal answer based on the mapping relationship.

[0012] Compared with the prior art, the beneficial effects of the present invention are: This invention proposes an intelligent question-and-answer system based on converter station operation and maintenance. It utilizes an intelligent interaction module to diversify answer presentation, facilitating quick user comprehension; an operation and maintenance knowledge module for centralized management of converter station operation and maintenance information; an intent understanding module to accurately interpret user-inputted operation and maintenance questions; and an answer generation module to precisely understand the intent behind the input question and quickly generate the optimal answer for converter station operation and maintenance. This invention can generate the optimal answer based on user-inputted questions, achieving intelligent solution and processing of converter station operation and maintenance related issues. It solves the problems of information fragmentation and reliance on human experience in existing converter station operation and maintenance systems, improving the efficiency and accuracy of operation and maintenance question-and-answer. Attached Figure Description

[0013] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only preferred embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0014] Figure 1This invention provides an overall structural schematic diagram of an intelligent question-and-answer system based on converter station operation and maintenance, as provided in an embodiment of the invention. Figure 2 A schematic diagram illustrating the process of constructing a converter station operation and maintenance knowledge graph according to an embodiment of the present invention; Figure 3 This is a schematic diagram of the knowledge data fusion processing flow of the operation and maintenance knowledge module provided in this embodiment of the invention; Figure 4 A flowchart illustrating the intent understanding module provided in an embodiment of the present invention; Figure 5 This is a flowchart illustrating the answer generation module provided in an embodiment of the present invention. Detailed Implementation

[0015] The principles and features of the present invention are described below with reference to the accompanying drawings. The listed embodiments are only used to explain the present invention and are not intended to limit the scope of the present invention.

[0016] Reference Figure 1 This embodiment provides an intelligent question-and-answer system based on converter station operation and maintenance. The system includes: Intelligent interaction module: Used to receive questions input by the user and present the best answer generated by the answer generation module based on the input question to the user; Operation and Maintenance Knowledge Module: Used to acquire knowledge data related to converter station operation and maintenance, and to build a converter station operation and maintenance knowledge graph; Intent understanding module: This module segments the user's input question into words and analyzes the intent behind the input question based on the word segmentation. Answer generation module: It receives questions processed by the intent understanding module, combines the questions with the converter station operation and maintenance knowledge graph to reason and generate the optimal answer.

[0017] In this embodiment, a standardized and structured converter station operation and maintenance knowledge graph is constructed through the operation and maintenance knowledge module. This integrates relevant converter station equipment ledger data, converter station operation and maintenance technical manuals, fault cases, and other data, solving the problems of fragmented, cumbersome, and difficult-to-pass-down traditional operation and maintenance knowledge. This provides accurate and authoritative knowledge support for intelligent question answering. The intent understanding module performs word segmentation and precise intent recognition on user-input questions, accurately capturing core needs and avoiding misunderstandings and irrelevant answers that are prone to occur in traditional manual Q&A, significantly improving the accuracy of question recognition. The answer generation module, combined with the structured operation and maintenance knowledge graph matching analysis, intelligently filters and generates the optimal Q&A results, replacing the traditional Q&A mode of manually reviewing materials and making experience-based judgments. This reduces labor costs while improving the accuracy and professionalism of Q&A. The intelligent interaction module builds a simple and efficient human-computer interaction channel, realizing the reception of questions and the visualization of answers. This makes operation convenient and lowers the operational threshold for converter station operation and maintenance Q&A.

[0018] The intelligent interaction module supports multiple interaction methods, including text input and voice input, to meet the usage habits and scenario needs of different users. For example, at the converter station maintenance site, maintenance personnel can quickly ask questions to the system through voice input, improving interaction efficiency; in the office environment, maintenance personnel can describe the problem in detail through text input.

[0019] The intelligent interaction module is also used to output the optimal answer generated by the answer generation module in the form of text, charts, and step-by-step instructions, supporting subsequent follow-up questions from the user. This includes automatically generating guiding follow-up questions to obtain key supplementary information when the user's input question information is incomplete; and recording user question-and-answer interaction logs. These logs record the user's input questions, answer results, and interaction process, and are used for the optimization of the intent understanding module.

[0020] Reference Figure 2 In the aforementioned operation and maintenance knowledge module, the construction of the converter station operation and maintenance knowledge graph specifically includes the following steps: S11. Collect various knowledge data related to converter station operation and maintenance, including converter station equipment ledger data, converter station operation and maintenance technical manual, and fault case data, and perform preprocessing operations on the collected knowledge data. S12. Divide the preprocessed knowledge data into structured data, semi-structured data, and unstructured data. Extract the first operational features from the structured and semi-structured data. Perform word segmentation and feature word extraction on the unstructured data. S13. Combine and analyze the first operation and maintenance features extracted in step S12 with the unstructured data after word segmentation and feature word extraction to construct a converter station operation and maintenance dictionary; S14. Based on the converter station operation and maintenance dictionary, determine the converter station equipment categories and their characteristics in various knowledge data, label the converter station equipment categories and their characteristics, and integrate the labeled converter station equipment categories and their characteristics into a training dataset. S15. Construct an operation and maintenance knowledge model. Input the training dataset into the operation and maintenance knowledge model for training. Use the trained operation and maintenance knowledge model to extract the second operation and maintenance feature from the unstructured data. Merge the second operation and maintenance feature with the first operation and maintenance feature to construct the converter station operation and maintenance knowledge graph.

[0021] Reference Figure 3 The operation and maintenance knowledge module is further used to fuse and process the acquired knowledge data, specifically including the following steps: S21. Based on the acquired knowledge data, extract the features of each entity in the knowledge data, and use the entity features to analyze the feature similarity between entities; S22. Based on the feature similarity between entities, match different entities to generate a potential matching set for entity matching; S23. Preset entity matching rules and filter the potential matching set according to the entity matching rules; S24. Based on the screening results of the potential matching set, establish matching relationships between different entities to eliminate knowledge redundancy in the converter station operation and maintenance knowledge graph.

[0022] The operation and maintenance knowledge module supports a combination of manual updates and scheduled automatic updates, and can be updated in real time based on new operation and maintenance data and expert experience, thus solving the problem of knowledge lag. The aforementioned operation and maintenance knowledge module solves the problem of fragmented operation and maintenance technical information, realizes centralized management of multi-source operation and maintenance data, and avoids the drawbacks of operation and maintenance technical information being scattered in different systems and documents and difficult to retrieve. By constructing a converter station operation and maintenance knowledge graph, scattered data is transformed into well-connected structured knowledge, which significantly improves knowledge retrieval and matching efficiency compared to traditional document-based storage.

[0023] Reference Figure 4 In the intent understanding module, the user-input question is segmented into words, and the intent of the input question is analyzed based on the segmentation. Specifically, this includes the following steps: S31. Use lexical analysis to segment the user's input question into words or phrases; S32. Initialize the word segmentation obtained in step S31 to obtain multi-dimensional word segmentation vectors, and construct the initialization matrix of the problem based on the multi-dimensional word segmentation vectors. S33. Model the initialization matrix using convolution, extract local features in the problem through convolution operations, and capture the correlation between word segments; S34. Introduce a convolutional neural network model to process the initialization matrix after convolutional modeling, extract the user's intent features, and analyze the user's final intent based on the intent features.

[0024] The intent understanding module is also used to analyze intent in user question-and-answer interaction logs. Figure 1 Addressing input issues and extracting meaning. Figure 1 Based on the grammatical features of the input problem, the intent understanding module is optimized to facilitate its ability to understand specific inputs. Figure 1 When the syntax of the input question is inconsistent with the input question, quickly determine the intent of the input question.

[0025] The intent understanding module can also adapt professional expressions through synonym replacement and other means to solve the shortcomings of existing general semantic understanding models in parsing operation and maintenance professional issues.

[0026] The intent understanding module, after preprocessing optimization such as word segmentation and combined with a convolutional neural network model, improves the accuracy of intent parsing while ensuring response speed, meets the real-time question-and-answer requirements of converter station operation and maintenance, and lays an accurate foundation for subsequent reasoning and decision-making.

[0027] Reference Figure 5 In the answer generation module, the system receives questions processed by the intent understanding module, combines the questions with the converter station operation and maintenance knowledge graph for reasoning, and generates the optimal answer. Specifically, this includes the following steps: S41. Based on the user's final intent, extract the entity from the question input by the user to obtain the question entity; S42. Match the problem entities with the converter station operation and maintenance knowledge graph to obtain the optimal candidate entity for each problem entity, and construct a problem entity set based on the optimal candidate entities; S43. Semantically vectorize the problem entity set and the converter station operation and maintenance knowledge graph, and perform semantic similarity matching between the user's final intent and the semantically vectorized problem entity set and knowledge graph to obtain a set of relationships; S44. Based on the user's final intent, the set of question entities, and the set of relationships, construct a mapping relationship with the knowledge graph, and deduce the optimal answer based on the mapping relationship.

[0028] The answer generation module overcomes the challenge of deep analysis of complex, multi-condition problems (such as concurrent consultations with multiple faults) by associating and matching the problem entity with the converter station operation and maintenance knowledge graph, thus making up for the inability of traditional rule-based reasoning to handle complex scenarios.

[0029] The optimal answer generated by the answer generation module is multi-dimensional and structured content. Compared with the single answer format of existing systems, it is easier for operation and maintenance personnel to quickly understand the faults of the converter station and the corresponding operation and maintenance methods, which is in line with the actual work needs of operation and maintenance personnel. The answer generation module mines the related knowledge in the data by mapping the final intent, the set of question entities and the set of relationships to the knowledge graph, realizes intelligent decision-making for operation and maintenance questions and answers, reduces the dependence on senior experts, solves the pain point of limited expert resources and difficulty in covering large-scale real-time question and answer needs, and helps novice operation and maintenance personnel to quickly solve problems.

[0030] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An intelligent question-and-answer system based on converter station operation and maintenance, characterized in that, The system includes: Intelligent interaction module: Used to receive questions input by the user and present the best answer generated by the answer generation module based on the input question to the user; Operation and Maintenance Knowledge Module: Used to acquire knowledge data related to converter station operation and maintenance, and to build a converter station operation and maintenance knowledge graph; Intent understanding module: This module segments the user's input question into words and analyzes the intent behind the input question based on the word segmentation. Answer generation module: It receives questions processed by the intent understanding module, combines the questions with the converter station operation and maintenance knowledge graph to reason and generate the optimal answer.

2. The intelligent question-and-answer system based on converter station operation and maintenance according to claim 1, characterized in that, The intelligent interaction module supports multiple interaction methods, including text input and voice input.

3. The intelligent question-and-answer system based on converter station operation and maintenance according to claim 2, characterized in that, The intelligent interaction module is also used to output the optimal answer generated by the answer generation module in the form of text, charts and step-by-step instructions, and to record the user question-and-answer interaction log. The user question-and-answer interaction log is used for the optimization of the intent understanding module.

4. The intelligent question-and-answer system based on converter station operation and maintenance according to claim 1, characterized in that, The construction of the converter station operation and maintenance knowledge graph specifically includes the following steps: S11. Collect various knowledge data related to converter station operation and maintenance, including converter station equipment ledger data, converter station operation and maintenance technical manual, and fault case data, and perform preprocessing operations on the collected knowledge data. S12. Divide the preprocessed knowledge data into structured data, semi-structured data, and unstructured data. Extract the first operational features from the structured and semi-structured data. Perform word segmentation and feature word extraction on the unstructured data. S13. Combine and analyze the first operation and maintenance features extracted in step S12 with the unstructured data after word segmentation and feature word extraction to construct a converter station operation and maintenance dictionary; S14. Based on the converter station operation and maintenance dictionary, determine the converter station equipment categories and their characteristics in various knowledge data, label the converter station equipment categories and their characteristics, and integrate the labeled converter station equipment categories and their characteristics into a training dataset. S15. Construct an operation and maintenance knowledge model. Input the training dataset into the operation and maintenance knowledge model for training. Use the trained operation and maintenance knowledge model to extract the second operation and maintenance feature from the unstructured data. Merge the second operation and maintenance feature with the first operation and maintenance feature to construct the converter station operation and maintenance knowledge graph.

5. The intelligent question-and-answer system based on converter station operation and maintenance according to claim 4, characterized in that, The operation and maintenance knowledge module is also used to integrate and process the acquired knowledge data, specifically including the following steps: S21. Based on the acquired knowledge data, extract the features of each entity in the knowledge data, and use the entity features to analyze the feature similarity between entities; S22. Based on the feature similarity between entities, match different entities to generate a potential matching set for entity matching; S23. Preset entity matching rules and filter the potential matching set according to the entity matching rules; S24. Based on the screening results of the potential matching set, establish matching relationships between different entities to eliminate knowledge redundancy in the converter station operation and maintenance knowledge graph.

6. The intelligent question-and-answer system based on converter station operation and maintenance according to claim 1, characterized in that, The intent understanding module specifically includes the following steps: S31. Use lexical analysis to segment the user's input question into words or phrases; S32. Initialize the word segmentation obtained in step S31 to obtain multi-dimensional word segmentation vectors, and construct the initialization matrix of the problem based on the multi-dimensional word segmentation vectors. S33. Model the initialization matrix using convolution, extract local features in the problem through convolution operations, and capture the correlation between word segments; S34. Introduce a convolutional neural network model to process the initialization matrix after convolutional modeling, extract the user's intent features, and analyze the user's final intent based on the intent features.

7. The intelligent question-and-answer system based on converter station operation and maintenance according to claim 6, characterized in that, The answer generation module specifically includes the following steps: S41. Based on the user's final intent, extract the entity from the question input by the user to obtain the question entity; S42. Match the problem entities with the converter station operation and maintenance knowledge graph to obtain the optimal candidate entity for each problem entity, and construct a problem entity set based on the optimal candidate entities; S43. Semantically vectorize the problem entity set and the converter station operation and maintenance knowledge graph, and perform semantic similarity matching between the user's final intent and the semantically vectorized problem entity set and knowledge graph to obtain a set of relationships; S44. Based on the user's final intent, the set of question entities, and the set of relationships, construct a mapping relationship with the knowledge graph, and deduce the optimal answer based on the mapping relationship.

Images